Speed up unit tests by removing sleep in start/stop torchserve

[mreso]

Description

This PR removes sleep(10) in our unit test utils and replaces it with a function which parses the log to wait the actual time it takes to start torchserve. This speeds up single unit tests by up to 4x.

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)

Feature/Issue validation/testing

Please describe the Unit or Integration tests that you ran to verify your changes and relevant result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• Run pytest test/pytest/test_example_micro_batching.py before and after change

Before:

After:

Checklist:

• Did you have fun?

[codecov]

Codecov Report

Merging #2383 (0d8281e) into master (e205e6b) will decrease coverage by 0.10%.
The diff coverage is 0.00%.

❗ Current head 0d8281e differs from pull request most recent head 24651ea. Consider uploading reports for the commit 24651ea to get more accurate results

@@            Coverage Diff             @@
##           master    #2383      +/-   ##
==========================================
- Coverage   72.13%   72.03%   -0.10%     
==========================================
  Files          78       78              
  Lines        3642     3647       +5     
  Branches       58       58              
==========================================
  Hits         2627     2627              
- Misses       1011     1016       +5     
  Partials        4        4              

Impacted Files	Coverage Δ
ts/arg_parser.py	25.80% <ø> (ø)
ts/model_server.py	0.00% <0.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[msaroufim]

Is the assumption that each new run of torchserve will overwite its logs?

It seems safe since tests are passing but just a bit worried we'll run into hard to debug testing issues later down the line

[mreso]

The tests still run sequentially and nothing is overwritten. This class only prints the logs even after TS initialization is completed. This replicates the same behavior as we had before where we ran TS in a background process and waited for 10 seconds. Now we parse the stdout pipe for the initialization completed signal and then just let this class print the rest of the log. This way folks can still see and debug issues as before.
EDIT: Wrong class, this class actually writes a log but the principle is the same. Previously we just piped the output into a file, now we first parse it and then write it into the file. All other behavior stays the same. From a dev perspective nothing changes besides that your unit test is not waiting 30+ seconds for nothing.
EDIT2: Okay, three time's a charm...the log file gets appended every times, so same behavior as ">>" vs ">". We could probably get the same result with a tee -a if you prefer that.